A Mean Shift-Based Pattern Formation Algorithm for Robot Swarms

Chase Vickery, Sayed Ahmad Salehi

Abstract: This paper proposes a distributed pattern formation algorithm for robot swarms that uses the mean shift algorithm as a central means for calculating robot movement. The proposed solution aims to allow for fast and scalable pattern formation. By representing the pattern as a discrete set of points on a coordinate plane, robots can use the mean shift algorithm to move towards nearby pattern points. The mean shift algorithm is further used to get robots to distribute around the pattern and prevent collisions among other robots by representing each neighboring robot as a negative pattern point that pushes robots away. In this way, the mean shift algorithm is used for both aggregation towards the pattern as well as regulating robot positions on the pattern. The simulation results presented demonstrate an ability of a swarm to form the pattern with a varying number of robots, showing scalability of the algorithm.